YAP1 and AR interactions contribute to the switch from androgen-dependent to castration-resistant growth in prostate cancer

The transcriptional co-activator Yes-associated protein 1 (YAP1), a key nuclear effector of the Hippo pathway, is a potent oncogene, and yet, the interaction between YAP1 and androgen receptor (AR) remains unexplored. Here we identify YAP1 as a physiological binding partner and positive regulator of AR in prostate cancer. YAP1 and AR co-localize and interact with each other predominantly within cell nuclei by an androgen-dependent mechanism in a hormone naive and an androgen-independent mechanism in castration-resistant prostate cancer cells. The growth suppressor MST1 kinase modulates androgen-dependent and -independent nuclear YAP1–AR interactions through directly regulating YAP1 nuclear accumulation. Disruption of YAP1 signalling by genetic (RNAi) and pharmacological (Verteporfin) approaches suppresses AR-dependent gene expression and prostate cancer cell growth. These findings indicate that the YAP1–AR axis may have a critical role in prostate cancer progression and serves as a viable drug target. Prostate cancer is regulated by androgen receptor signalling; however, progressive disease can also be androgen-independent, thus hindering antiandrogen therapy. Here, interaction of the Hippo pathway protein, YAP1, with the androgen receptor is shown to confer castration-resistant prostate cancer.

P rostate cancer (PC) is a leading cause of cancer deaths among men in the Western countries 1 . Aberrant and deregulated androgen receptor (AR) signalling is a potent promoter of PC development, progression and metastasis 2,3 . AR gene amplification 4 , and mutations 5 that increase or decrease sensitivity and/or specificity to its ligands 6 , oncogenic growth factor signalling 7 , and altered AR co-regulators 8 have been shown to cause an aberrant AR activation, even in the presence of very little androgens in circulation 9,10 . Therefore, the antiandrogen therapy has some survival benefits for patients with advanced PC, but this strategy is temporary because the metastatic castrationresistant prostate cancer (CRPC) emerges. Metastatic CRPC is lethal because there is no effective therapy for it. Despite recent advances [10][11][12][13] , the molecular mechanisms contributing to invasive CRPC are poorly understood. Lines of evidence suggest that AR is a key driver of this lethal disease, even in the presence of enzalutamide 14 , the second-generation potent inhibitor of AR, but the mechanism of how AR regains its functions and derives metastatic CRPC remains obscure.
AR is a transcription factor and a member of the steroid hormone receptor superfamily 15 . AR has three major functional domains: an NH2-terminal transactivation domain (NTD) that mediates a ligand-dependent or ligand-independent activation of AR 16 , a DNA binding domain (DBD) that interacts with cis-DNA, called the androgen response element (ARE), and a COOH-terminal ligand-binding domain (LBD) that mediates dimerization and a ligand-dependent transactivation of AR 17 . Studies based on cell models-mostly in COS7 cells-indicate that unliganded and structurally intact AR normally resides within the cytoplasm as an inactive form through interaction with heat shock proteins (HSPs) such as hsp90 18 . On ligand binding, AR disassociates from hsp90 and then enters into cell nuclei, where it binds to the ARE and regulates its target gene expression through interaction with co-regulatory proteins (that is, co-activators or co-repressors) 17,19 . Altered expression of the AR co-regulators that comprise scaffold proteins with or without chromatin modifying functions has been implicated in metastatic CRPC 8,20,21 . Since AR-targeted therapy has limited clinical benefits, there is a great deal of interest in targeting the AR regulators to improve current therapies. YAP1 (Yki in Drosophila) is a transcriptional co-activator and regulates diverse cellular biology including growth 22 , apoptosis 23 , differentiation 24 , cell-cell interactions 24,25 , epithelial to mesenchymal transition (EMT) 26,27 , oncogenic cell transformation 28 and chemoresistance 29,30 . YAP1 interacts with various transcription factors such as TEF/TEAD 31 , SMADs 32 or TBX5 33 to exert its biologic functions in the cell. MST1/2 and LATS1/2, core kinase components of the Hippo pathway are potent negative regulators of YAP1. The cascade of MST1/2 and LATS1/2 signalling was demonstrated to inactivate YAP1 by promoting its inhibitory Ser127 (S127) site phosphorylation, cytoplasmic retention and proteasome-mediated degradation 34 . Activation of YAP1 due to the gene amplification or the loss-offunction of MST1/2 or LATS1/2 is linked to the aetiology of many cancers including lung 35 , colon 36 , ovarian 37 , head and neck 38 , liver 39,40 , meningioma 41 , thyroid 42 and stomach 43 . In addition, emerging evidence suggests that YAP1 may also play a critical role in the pathobiology of PC 44 . However, the role and mechanism of YAP1 in PC remains to be explored.
In this study, we investigate the biochemical and functional association between YAP1 and AR in PC. We demonstrate that YAP1 and AR co-localize and form protein complexes primarily in cell nuclei and that the complex formations between the two proteins are androgen-dependent in castration-sensitive (CS), but are androgen-independent (AI) in CR PC cells. In addition, we show that MST1 is a key negative regulator of YAP1-AR interactions, which may play crucial role in AR-dependent gene expression and PC cell growth in vitro and in vivo. Our study identifies new functions of the YAP1 interaction with AR.

YAP1 and AR interact with each other in prostate cancer cells.
To explore the relationship between YAP1 and AR, we utilized PC clinical samples and cell models and employed cellular, molecular and biochemical approaches. In this study, wellestablished CS or hormone naive LNCaP and CR or hormone refractory C4-2 PC cells were extensively used because they are genetically related and express functional AR, but differentially respond to androgen hormone signalling for their growth in vitro and in vivo 45 . The C4-2 cell line was generated by in vivo selection of LNCaP cells recycling through castrated mice 45 .
First, we analysed the publicly available YAP1 data using the www.cbioportal.org online platform. The analysis indicated that unlike other cancers such as ovarian or cervical cancer, about 3-6% of PC cases showed genetic alterations in YAP1 gene (that is, deletion or amplification; Fig. 1a). Second, we assessed the expression of YAP1 protein in the histologic sections of normal prostate (NP) and PC tissues by immunohistochemistry (IHC). YAP1 protein was abundantly stained in NP and PC tissues (Fig. 1b, middle and right panels, respectively). Notably, staining of YAP1 protein, which was predominantly accumulated in cell nuclei, was not uniform and showed heterogenic features (that is, overstaining, under-staining or no staining) within the same samples and amongst the cases. IgG control did not visualize YAP1 protein expression in the tissue (Fig. 1b, left panel), indicating that YAP1 staining was specific.
To determine whether native YAP1 and AR biochemically interact with each other in vivo, we performed co-immunoprecipitation (co-IP) and western blot (WB) experiments utilizing total lysate isolated from fresh-frozen normal and cancerous human prostate tissue specimens. The results showed that endogenous YAP1 and AR formed protein complexes ( Fig. 1c; Supplementary Fig. 1a). Similarly, co-IP and WB experiments demonstrated that YAP1 and AR interacted with each other in total lysate obtained from LNCaP and C4-2 cells that were grown in serum-fed conditions ( Fig. 1d; Supplementary Fig. 1b). The interactions between the two proteins were specific because IgG control did not show any visible interaction (Fig. 1c,d). In addition, androgen (dihydrotestosterone: DHT) exposure increased YAP1-AR interactions four-fold in LNCaP cells compared with vehicle (EtOH) control (Fig. 1e, lane 2 versus lane 1; Supplementary Fig. 1c). Surprisingly, however, YAP1-AR interactions occurred independently of DHT exposure in C4-2 cells (Fig. 1e, lane 3; Supplementary Fig. 1c). Although the C4-2 cell line does not require androgens for growth and survival 45 , it still responds to androgen hormone signalling due to the expression of functional AR. For this reason, YAP1-AR interactions were further enhanced by DHT exposure in C4-2 cells when the result was normalized to total YAP1 protein (Fig. 1e, lane 3 versus lane 4; Supplementary Fig. 1c).
YAP1 and AR interactions occur in cell nuclei. To identify the subcellular location where YAP1 and AR primarily interact and whether that is regulated by androgen, we performed co-IP and WB experiments utilizing cytoplasmic and nuclear fractions isolated from LNCaP cells at 0, 4 and 24 h post DHT treatment. Co-IP and WB showed that YAP1 and AR interacted predominantly in cell nuclei and that androgen exposure increased AR presence in the YAP1 immune complex in a time-dependent manner ( Fig. 2a; Supplementary Fig. 2a). Here we want to clarify that AR appears to be equally distributed between the cytoplasmic ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms9126 and nuclear fractions without androgen exposure (Fig. 2a, lane 1 and 2, input). One possible explanation for this is that the AR in LNCaP cells has a Thr877Ala mutation, and this mutation has been implicated to favour AR nuclear localization in the absence of androgen 46 . Nevertheless, on androgen exposure, the majority of AR localizes to the cell nuclei, as demonstrated by WB (Fig. 2a, lane 3 versus 4 and 5 versus 6, input) and immunofluorescence imaging ( Fig. 2c left panel versus right panel). Thus, our data are internally consistent and aligned with the published literature 46 . In addition, androgen exposure increased the abundance of YAP1 in the AR immune complex compared with vehicle exposure ( Supplementary Fig. 2b).
Nevertheless, unlike LNCaP, the interaction between YAP1 and AR in C4-2 cells was independent of androgen exposure, as shown by co-IP and WB experiments ( Fig. 2b; Supplementary  Fig. 2c). In addition, the protein complex formation between YAP1 and AR were reduced to undetectable levels in the nuclei of C4-2 cells with the RNA interference (RNAi)-mediated depletion of YAP1 ( Supplementary Fig. 3a), indicating that YAP1 and AR interactions are specific. Moreover, co-immunofluorescence experiments demonstrated that compared with vehicle control, YAP1 and AR proteins co-localized in LNCaP cells in androgendependent manner ( Fig. 2c; Supplementary Fig. 3b), whereas the co-localization was AI in C4-2 cells ( Fig. 2d; Supplementary  Fig. 3c). Taken together, the cell nuclei are likely the primary interaction site of YAP1 and AR proteins.
YAP1 and AR interactions are resistant to enzalutamide. To determine the impact of a potent AR inhibitor enzalutamide 47 on YAP1-AR interactions, we repeated co-IP and WB experiments utilizing cytoplasmic and nuclear fractions isolated from LNCaP and C4-2 cells that were exposed to vehicle or enzalutamide with or without DHT. The results showed that enzalutamide attenuated androgen-dependent and AI interactions between nuclear YAP1 and AR proteins in LNCaP cells ( Fig. 3a; Supplementary Fig. 4a), but failed to block the nuclear interactions between two proteins in C4-2 cells ( Fig. 3b; Supplementary Fig. 4b). Accordingly, unlike LNCaP, enzalutamide also failed to attenuate the growth of C4-2 cells in vitro, regardless of androgen exposure (Fig. 3c).
YAP1 and AR interactions are regulated by MST1. Previously, we reported that C4-2 cells express significantly lower levels of MST1 than LNCaP cells 48 . Results from this report and the above findings suggest that MST1 is a potent negative regulator of the YAP1-AR interaction. To test this possibility, we performed co-IP and WB experiments utilizing cytoplasmic and nuclear fractions obtained from LNCaP cells with or without MST1 knockdown plus or minus DHT treatment. Knockdown of MST1 markedly increased androgen-dependent and AI interactions between YAP1 and AR compared with mock controls ( Fig. 3d; Supplementary Fig. 5a). Under the same experimental conditions above, MST1 depletion increased YAP1 nuclear localization with or without androgen exposure ( Fig. 3d; Supplementary Fig. 5a). Accordingly, knockdown of MST1 promoted the androgendependent and AI growth of LNCaP cells in vitro (Fig. 3e), mimicking the functional behaviours of C4-2 cells (see Fig. 3b,c). Conversely, a controlled expression of the ectopic MST1 suppressed the interaction between YAP1 and AR proteins in C4-2 cells ( Fig. 4a; Supplementary Fig. 5b) and inhibition of YAP1-AR interactions coincided with reduction in nuclear abundance of YAP1 (Fig. 4b,  ARTICLE promoter DNA methylation was implicated in PC 50 . In agreement with this observation, our analysis indicated that expression of LATS mRNA and protein was very low in LNCaP and C4-2 in comparison with that of the positive control HeLa or C2C12 cell lines ( Supplementary Fig. 6a,b), as demonstrated by PCR and WB, respectively. These observations led us to believe that MST1 could regulate YAP1, independently of LATS. To test this hypothesis, we performed a series of biochemical assays. First, co-IP and WB showed that native MST1 and YAP1 formed protein complexes in LNCaP cells ( Fig. 4d; Supplementary  Fig. 7a). Second, in vitro kinase assays revealed that the MST1 immune complex that was precipitated from LNCaP cells was capable of phosphorylating the recombinant GST-YAP1 fusion peptide (residues 2-150) comprising the S127 phosphorylation site ( Fig. 4e; Supplementary Fig. 7b). Third, GST pull-down and in vitro kinase assays demonstrated that recombinant, the preactivated MST1 kinase interacted with and phosphorylated the GST-YAP1-S127 peptide ( Fig. 4f; Supplementary Fig. 7c). There was no detectable interaction or phosphorylation signal between MST1 and the GST only (control) peptide under these experimental conditions, indicating that the observation was specific. Fourth, more importantly, unlike LATS1/2, knockdown of MST1 reduced phospho-YAP1-S127 levels by 60% in C4-2 cells compared with mock control (Supplementary Fig. 13). These findings consistent with our notion suggest that MST1 is a potent direct negative regulator of YAP1 nuclear localization. This may be a mechanism by which MST1 negatively regulates YAP1-AR interactions.
YAP1-WW/SH3 domain interacts with AR. YAP1 protein consists of several functional domains including proline-rich (PR), TEAD, WW, SH3, coiled-coil (CC), transactivation domain (TAD) and PDZ domains (Fig. 5a). To map the YAP1 domain that facilitates the interaction with AR, we generated GST-YAP1-WT and GST-YAP1 truncation mutant constructs comprising PR and TEAD domains (GST-YAP1-PR/TEAD (residues 2-150), WW and SH3 domains (GST-YAP1-WW/SH3; residues 151-296) or CC, TAD and PDZ domains (GST-YAP1-CC/TAD/ PDZ; residues 297-504) and attempted to express them in bacteria as a recombinant protein (Fig. 5b, Coommassie blue stain; Supplementary Fig. 8a). Due to technical difficulties, we were unable to express the full-length YAP1 as a GST fusion protein at all, even with several attempts. Luckily, we would be to successfully express YAP1-PR/TEAD or YAP1-WW/SH3 domain as a GST fusion peptide. In addition, we managed to express GST-YAP1-CC/TAD/PDZ fragment, though the yield was much lower than GST-YAP1-PR/TEAD and GST-YAP1-WW/SH3 peptides. To our knowledge, this is the first report showing the expression of YAP1 as a GST fusion protein in bacteria.
To identify the YAP1 domain that potentially interacts with AR, we performed GST-pull-down experiment utilizing recombinant GST only (negative control), GST-YAP1-PR/TEAD, GST-YAP1-WW/SH3 or GST-YAP1-CC/TAD/PDZ fusion peptide as a bait and total lysates isolated from LNCaP cells, which was used as a source of AR ( Fig. 5b; Supplementary  Fig. 8a). Although no interaction was observed between GST-YAP1-PR/TEAD and AR (Fig. 5b, lane 3), there was an apparent interaction between GST-YAP1-WW/SH3 and AR (Fig. 5b, lane 4). The observed interaction was specific because GST only peptide did not show any interaction (Fig. 5b, lane 2). Since the expression of GST-YAP1-CC/TAD/PDZ in bacteria resulted in a very low yield, we were unable to detect visible interactions between GST-YAP1-CC/TAD/PDZ and AR under these experimental conditions (Fig. 5b,  play a key role in its interaction with YAP1, because YAP1 interacted with both naturally occurring short (N-terminal) and the full-length (FL) AR forms. To test this hypothesis, we performed co-IP and WB experiments using total lysate obtained from C4-2 cells that were transiently transfected to express HA-tagged vector (control), AR-FL or AR-NTD truncation mutant ( Fig. 5c; Supplementary Fig. 8b). The results showed that ectopic AR-FL or AR-NTD truncation mutant formed protein complexes with endogenous YAP1 (Fig. 5c, lane 2, 3, respectively). The HA-vector produces a very small peptide, which cannot be detected under these experimental conditions, possibly due to running off the 8% protein gel or degradation, did not show any apparent interaction with native YAP1 (Fig. 5c, lane 1; Supplementary Fig. 8b, lane 1). Thus, these findings further strengthen the specificity of interactions between YAP1 and AR and demonstrate that NTD of AR most likely provides a docking site for YAP1.
YAP1 regulates the expression of AR-responsive genes. To test the impact of YAP1 on AR responsive gene expression, we conducted prostate-specific antigen (PSA) promoter-driven luciferase reporter (p61-Luc) assays. PSA is a well-characterized AR-dependent gene and widely used to assess AR transactivation functions. LNCaP cells were transiently co-transfected with p61-Luc reporter along with mock (control), YAP1-WT (residues 2-504), YAP1-DN (residues 58-504) or YAP1-DC (residues 2-290) expression construct, followed by treatment with vehicle or DHT. Transient expression of YAP1-WT increased the androgen induction of PSA promoter reporter activation about threefold compared with mock vector (Supplementary Fig. 9a; Fig. 5d, respectively). Ectopic expression of YAP1-DN did not significantly affect androgen-induced PSA promoter activation, whereas YAP1-DC lacking its transactivation domain failed to enhance the androgen induction of PSA promoter activation. This indicates that C-terminal TAD domain of YAP1 plays key role in AR transactivation. In addition, quantitative PCR demonstrated that transient knockdown of YAP1 with two different short hairpin RNA (shRNA) significantly reduced the endogenous expression of other well-characterized AR-dependent genes KLK3, PSMA, FKBP5 and TMPRSS2 at mRNA levels relative to control shRNA (Fig. 5e). Moreover, transient overexpression of YAP1-WT significantly increased the growth of LNCaP cells in vitro, regardless of androgen exposure in comparison with vector control (Supplementary Fig. 9b). These observations suggest that YAP1 is a prominent regulator of AR-dependent signalling. ARTICLE knockdown (shYAP1) or control (shControl) C4-2 cell models with the gene specific shRNAs ( Fig. 6a; Supplementary Fig. 10a).
Utilizing these cells, we performed series of biological assays. Knockdown of YAP1 decreased C4-2 cell growth in vitro in a time-dependent manner (Fig. 6b). Similarly, YAP1 knockdown by siRNA also attenuated the growth of LNCaP cells in vitro ( Supplementary Fig. 10b). Likewise, YAP1 knockdown suppressed the androgen induction of C4-2 cell growth in monolayer ( Supplementary Fig. 10c) and in three-dimensional (3D) cultures in comparison with controls ( Fig. 6c). Cell cycle analysis showed that a cell cycle arrest at G0/G1 most likely caused growth retardation by YAP1 knockdown (Supplementary Fig. 10d). In addition, YAP1 knockdown significantly diminished C4-2 cell invasion in Matrigel-coated Transwell chambers in a timedependent manner (Fig. 6d). Besides, stable YAP1 knockdown attenuated the expression of AR target genes in vitro ( Supplementary Fig. 10e), which was in agreement with the data in Fig. 5e. Alternatively, we repeated the above functional assays utilizing a small molecule inhibitor, verteporfin (VP). VP was identified as a potent inhibitor of YAP1 (ref. 51). VP exposure attenuated the growth of C4-2 cells in dose-and time-dependent manners (Fig. 7a,b). VP also suppressed the androgen induction of LNCaP or C4-2 cell growth ( Supplementary Fig. 11a-c). Similarly, VP also suppressed the growth of AR-negative PC3 and DU145 cells in a similar manner ( Supplementary Fig. 11d). However, PC3 and DU145 cells were relatively resistant to VP compared with LNCaP and C4-2 cells. VP also caused apoptosis in a dosedependent manner ( Fig. 7c; Supplementary Fig. 11e) and cell cycle arrest at G0/G1 (Supplementary Fig. 11f). Likewise, VP profoundly attenuated the ability of C4-2 cells to form spheres in Matrigel (Fig. 7d) and to invade through Matrigel-coated Transwell chamber (Fig. 7e).
To gain more insights into how VP suppressed PC cell growth, we conducted immunofluorescence, co-IP and WB experiments to see if VP exposure impairs YAP1 nuclear localization and its interaction with AR. As proposed, VP exposure suppressed YAP1 nuclear localization (Fig. 7f) and YAP-AR interactions ( Fig. 7g; Supplementary Fig. 12a), without affecting AR protein expression or its nuclear localization relative to control (Fig. 7g, input; Supplementary Fig. 12b, respectively). In agreement with these observations, decreases in YAP1-AR interactions by VP were accompanied with inhibition of AR-dependent PSA protein expression (Fig. 7g, input; Supplementary Fig. 12c). Moreover, WB and immunofluorescence experiments showed that unlike vehicle control, VP exposure increased phospho-YAP1-S127 protein levels (Fig. 7g, input; Supplementary Fig. 12d,e, respectively).
To determine whether increases in YAP1 phosphorylation by VP correlates with the induction of Thr183 (T183) phosphorylation, a key mediator of MST1 kinase activity 52 , we assessed the levels of total and phospho-YAP1-S127 along with total and phosho-MST1-T183 by WB in C4-2 cells. Relative to control, VP exposure increased phospho-MST1-T183 levels that coincided with upregulation of phospho-YAP1-S127 (Fig. 7g, input;  Supplementary Fig. 12f, indicating that VP promotes YAP1-S127 phosphorylation by activating MST1). In addition, compared with control siRNA, transient transfection of LATS1/ 2 siRNA did not alter the VP induction of phosho-YAP1-S127-, whereas transient transfection of MST1 siRNA caused a significant reduction in YAP1-S127 phosphorylation by VP ( Supplementary Fig. 13). Moreover, VP significantly decreased the expression of AR-responsive genes KLK3, FKBP5, and TMPRSS2 without affecting AR protein levels ( Supplementary  Fig. 14a, Fig. 7g, input, respectively). These observations suggest that MST1 is a potent direct regulator of YAP1 under conditions where LATS1/2 expression or functions is lost.

YAP1 silencing suppresses prostate tumour xenografts.
To determine the biological significance of the above findings in vitro, we performed mouse xenograft experiments. Luciferasetagged shControl or shYAP1 C4-2 cells were subcutaneously injected into immunodeficient male mice (n ¼ 10 per condition) in the right and left flanks. Tumour growth was monitored for up to 5 weeks by luciferase imaging in live mice. Compared with shControl, YAP1 knockdown significantly reduced the number and size of tumours formed by shYAP1 C4-2 cells (Fig. 8a,b). To see whether YAP1 knockdown was able to suppress AR-dependent gene expression in vivo, we assessed the levels of PSA protein by WB along with YAP1 and AR in shControl and shYAP1 xenograft tissues (n ¼ 4 per group). YAP1 silencing significantly reduced PSA protein levels without affecting AR protein levels relative to shControl ( Supplementary Fig. 14b,c).
Similarly, IHC analysis of AR and YAP1 expression in xenograft tissues sections further supported the WB data (Fig. 8c).
To determine whether suppression of tumour growth correlates with an impaired cell proliferation and induction of apoptosis, we evaluated the expression of Ki-67 proliferation and cleaved caspase 3 (c-Cas3) apoptotic markers in xenograft tissues by IHC. The results showed that a number of proliferative (Ki-67 positive) cells were reduced, while a number of apoptotic (c-Cas3 positive) cells were increased in shYAP1 C4-2 tumours compared with shControl (Fig. 8c). In addition, permanent YAP1 knockdown attenuated the expression of KLK3, PSMA, FKBP5 and TMPRSS2 mRNA in shYAP1 C4-2 xenografts relative to shControl (Fig. 8d). Taken together, these observations consistently demonstrate that the YAP1-AR axis is a key mediator of PC cell growth and survival, which can be negatively regulated by upstream MST1 signalling (Fig. 8e).

Discussion
In the current study, we have demonstrated that a nuclear interaction between YAP1 and AR that is regulated by MST1 signalling, possibly via a LATS1/2 bypass mechanism, may play a prominent role in the emergence of advanced PC. We provide convincing evidence to support this conclusion. Our data demonstrate that unlike CSPC cells, YAP1-AR interactions are AI and resistant to enzalutamide in CRPC cells. Our data also demonstrate that knockdown of MST1 in CS LNCaP cells increases YAP1 nuclear localization and YAP1-AR interactions, which coincided with augmented cell growth, independently of androgen exposure, and that the controlled expression of ectopic MST1 had the opposite effect. In addition, disruption of YAP1 signalling by a genetic or pharmacological approach caused the suppression of C4-2 cell growth in vitro or xenografts in vivo as well as inhibition of AR-dependent gene expression. These findings suggest that the YAP1-AR axis is a promising cancer drug target.
The Hippo pathway regulates tumorigenesis and organ size by restricting cell proliferation and promoting apoptosis. MST1,   56 showed that amplification of YAP1, which is uncommon in PC (Fig. 1a), was sufficient to promote cell survival in the animal model of pancreatic cancer and that these functions of YAP1 might occur independently of MST1/2. Nevertheless, genetic studies in mice showed that liver cells with LATS depletion were able to partially maintain their phospho-YAP1-S127 status 52 , suggesting that other kinases regulate YAP1 independently of LATS. Herein, we identified MST1 as a potent direct regulator of YAP1. Our data showed that MST1 might phosphorylate YAP1, possibly through LATS bypass mechanism in LNCaP or C4-2 PC cells; these cells naturally express very low or undetectable LATS protein or in which expression of LATS was depleted by siRNA. Notably, loss of LATS expression through DNA methylation was documented in PC 47 and other cancers 51 . Altogether, MST1 and LATS sequentially or independently regulate YAP1 and these mechanisms, perhaps, occur in a context-dependent or cell-specific manner. In addition, protein kinase B/AKT1 was suggested to phosphorylate YAP1 on S127 (ref. 57). Previously, our laboratory reported that MST1 functionally interacted with AKT1 (refs 52,58,59). In the current study, however, we did not examine the involvement of AKT1 in the regulation of YAP1 by MST1. Nevertheless, it is unlikely that AKT1 interferes YAP1 phosphorylation in LNCaP and C4-2 cells because these genetically-related cells express constitutively active AKT1 due to the loss of PTEN (phosphatase and tensin homologue) 60 ; PTEN is a key negative regulator of PI-3-Kinase-AKT1 signalling 61 . However, our data revealed that YAP1 nuclear accumulation and its interaction with AR were significantly different in LNCaP and C4-2 with or without androgen exposure ( Fig. 1d; Fig. 2). These differences appear to correlate with MST1 activity, but not with the activity of LATS or AKT1 ( Fig. 1d; . Besides, Jiang et al. 44 identified additional phosphorylation sites (S163/164 and T63) on YAP1 in the lysate of LNCaP xenografts through a quantitative proteomic approach; however, that study did not show a change in phospho-YAP1-S127 levels. One possible explanation for it was that the MST1 kinase activity might be inactivated in LNCaP xenografts, possibly due to the expression of the constitutively active AKT1, which is known to inactivate MST1 via phosphorylation 59 . Thus, it would be interesting to know in a future investigation whether MST1 and LATS sequentially or independently regulate phosphorylation of S163/164 and T63 and whether these phosphorylation sites alter YAP1 nuclear localization and YAP1-AR interactions in the context of CRPC cells.
Lines of evidence have indicated that overexpression or nuclear accumulation of YAP1 may be associated with poor cancer prognosis 23,40,[62][63][64][65] . Our data consistent with the literature have also suggested that nuclear YAP1 plays a critical role in the aetiology of CRPC by interacting with AR. Here we showed that androgen increased YAP1 nuclear localization and YAP1-AR interactions in CS LNCaP cells. Nevertheless, these molecular events cannot be reversed by androgen depletion in CR C4-2 cells, which express about 40-50% less MST1 than LNCaP 48 . Our laboratory previously reported that reduction or loss of MST1 might play a prominent role in PC progression 48,52 . Herein, we showed that loss of MST1 functions strongly correlated with increases in AI interactions of YAP1 and AR proteins in CRPC cells and that knockdown of MST1 in CS LNCaP cells promoted androgen-dependent and AI YAP1-AR interactions ( Fig. 1e;  Fig. 2b; Fig. 3f). The interaction between the two proteins is biologically relevant because YAP1 knockdown prevented AR-dependent gene expression and PC cell growth in vitro (Figs 5,6) and xenografts in vivo (Fig. 7). Nonetheless, future investigations are necessary to explore the underlying mechanism of how changes in YAP1 functionality alters the genome-wide AR-DNA interactions and the AR transcriptional programs relevant to metastatic CRPC. Moreover, AR variants, lacking LBD and maintaining NTD and DBD, are implicated in the pathobiology of metastatic CRPC 14 . AR variants appear to be commonly observed in PC following androgen-deprivation therapy or exposure to antiandrogens 14 . Since antiandrogen therapy aimed at blocking ligand-dependent AR activity has limited clinical benefit, development of new therapies targeting AR-NTD may help to overcome this limitation. Herein, we showed that YAP1 interacted with both full length and naturally occurring truncated short AR forms independently of androgen exposure in CRPC cells (Fig. 2). In addition, YAP1 was identified as a key regulator of stem cell maintenance including cancer stem cells and EMT 26 . Mechanistically, cancer stem cell and EMT are believed to play a prominent role in metastatic CRPC. In addition, activation of YAP1 is implicated in the cellular response to DNA damage 23,62,64,66 and resistance to chemotherapeutics 40,65 . YAP1 activation has also been shown to be an essential bypass mechanism in K-Ras-dependent tumours when K-Ras signalling is inhibited in pancreatic, colon and lung cancer 26,56 . In addition, overexpression of YAP1 has recently been reported to play a role in castration resistance 67 . Thus, blocking of the AR-YAP1 signalling axis may have important therapeutic implications in metastatic CRPC.
VP is an FDA-approved drug prescribed for age-related macular degeneration in photodynamic therapy (PDT) 68 . A study by Liu-Chittenden et al. 51 showed that VP could bind the purified YAP1 protein and attenuate the interaction between YAP1 and TEAD in vitro, which correlated with suppression of YAP1-dependent liver cancer in mice 51 . Our study further clarified a mode of action of VP on YAP1. We showed that VP hindered YAP1 nuclear translocation by increasing its inhibitory S127 site phosphorylation by activating MST1. In addition, VP suppressed the growth of enzalutamide-resistant C4-2 cells that coincided with the inhibition of YAP1-AR interactions (Fig. 3). Ironically, AR-negative PC3 or DU-145 PC cells displayed resistance to VP induced growth retardation relative to AR positive LNCaP or C4-2 cells. Taken together, VP shows a great promise for the treatment of AR-dependent PC with the potentials of conferring enzalutamide resistance. Nonetheless, future investigations are warranted to evaluate the therapeutic efficacy of VP in metastatic CRPC in preclinical and clinical settings. In summary, our data demonstrate that increases in nuclear YAP1 and AR interactions, possibly due to the loss of Hippo functions, may play a critical role in metastatic, lethal PC progression. The lysates from LNCaP cells were used as a source of AR. After extensive washing of unbound proteins, bound protein eluted and analysed by 8% sodium dodecyl-PAGE (SDS-PAGE). Coomassie blue staining visualized the GST only or GST-YAP1 fusion peptides. Western blots were employed to visualize the bound AR. In vitro kinase assay was performed as described previously 52 . Briefly, GST only, GST-YAP1 (2-150)-S127 wild-type and GST-YAP1 (2-150)-S127A mutant fusion peptides were used as substrates to perform the non-radioactive kinase assays using the pre-activated, recombinant, full length, human MST1 protein kinase (Millipore, 14-624).
Cell engineering and cell viability assays. Lentivirus carrying control shRNA or YAP1 shRNA expression construct in combination with pCMV-VSV-G and pCMV-dR8.2 dvpr packaging plasmids were transfected into HEK-293T cells using Lipofectamine 2000 transfection reagent according to the manufacturer's instructions (Invitrogen) and as described 52 . Lentiviral particles were collected at 48 h post transfection and stored at À 80°C until use. For stable cell engineering, 3 Â 10 5 C4-2 cells grown in six-well plates were infected with lentiviral particles carrying control shRNA or YAP1 shRNA. Puromycin (1 mg ml À 1 ) resistant shRNA control or shRNA YAP1 cells were selected and amplified for the future biological assays. Stable control shRNA or YAP1 shRNA C4-2 cells were also engineered to express luciferase reporter for animal experiments. Cell viability was assessed in 96-well culture dish under varying conditions utilizing a MTS CellTiter 96 AQueous assay system according to the manufacturer's instruction (Promega; Madison, WI) and as described 59 .
RNA isolation and quantitative PCR. Total RNA from cells or tumor xenografts were extracted suing RNeasy RNA isolation kit according to the manufacturer's instructions (Qiagen; Maryland, MA). Complementary DNA (cDNA) was synthesized using total RNA (2 mg per reaction) with SuperScript II reverse transcriptase and oligo (dT)-12 primers (Invitrogen; Grand Island, NY). Quantitative PCR (qPCR) was carried out using Fast SYBR Green Master Mix (Invitrogen) and Applied Biosystems 7500 Real-Time PCR system (Life Technologies). RNA expression changes were determined using a 2-DCt method 70 . GAPDH mRNA was used as an internal control in all qPCR reactions. Supplementary Table 1 shows the qPCR primers used for YAP1, KLK3, PSMA, FKBP5, TMPRSS2 and GAPDH mRNA amplifications.
Cell cycle analysis and apoptosis assays. Cells were treated either with DMSO (vehicle) or Verteporfin (SML0534; Sigma-Aldrich) for 48 h in serum-fed conditions. Cells were fixed in 0.5% paraformaldehyde for 1 h at 4°C, washed with PBS, and then permeabilized in 70% cold EtOH for 1 h at 4°C. Cells were washed with cold PBS and 1 Â binding buffer containing 50 ml of propidium iodide (PI) staining solution (Cat#00-6990, eBioscience,) was added and incubated at 37°C for 30 min. Cells were filtered through a 0.45-mm filter and analysed by flow cytometry (CyAn ADP Analyzer, Beckman Coulter). For Annexin V staining, an Annexin V Apoptosis Detection Kit APC (Cat#88-8007, eBioscience,) was used. Briefly, cells were incubated with APC-conjugated Annexin V for 15 min at room temperature, washed with binding buffer, and then incubated with PI Staining Solution for 3 h at 4°C and kept in the dark until analysis.
Sphere formation assays. Sphere formation assays were performed as described 59 . Briefly, control shRNA and YAP1 shRNA cells (500 cells per well) suspended in 100 ml ice-cold Matrigel in RPMI medium (1:1 ratio) were overlaid onto the presolidified 50% Matrigel in 24-well plates (100 ml per well). Cells were fed with 500 ml RPMI medium containing 10% FBS and grown for 14 days with a change of medium every 3 days. For the VP study, C4-2 cells (500 cells/well) were grown on Matrigel and treated either with DMSO (vehicle) or VP in serum-fed conditions for 14 days. Spheres were imaged and then manually quantified.
Invasion assays. Control shRNA or YAP1 shRNA cells (1 Â 10 5 cells per well) were seeded in a 24-well Boyden chamber with an 8-mm filter coated with 20% growth factor reduced Matrigel. Cells were grown in RPMI medium containing 10% FBS for 16, 24 and 48 h in a 37°C cell culture incubator supplemented with 5% CO 2 . Cells in the inner side of the chamber were gently removed by scraping with a wet cotton swab. Invaded cells at the outer side of the chamber were fixed with 4% formaldehyde for 30 min at room temperature and rinsed twice with PBS. Cells were stained with 0.5% Crystal Violet for 20 min at room temperature and then rinsed with tap water to remove excess dye. Five random fields of stained cells were imaged using bright field microscopy at 20 Â magnification and average cell numbers per field were plotted as a function of time.
Animal experiments. Xenograft experiments in mice were conducted as previously described 52 and according to a protocol approved by the Animal Care and Use Committee (IACUC) at Cedars-Sinai Medical Center. Briefly, shRNA control or shRNA YAP1 C4-2 cells (1 Â 10 6 cells per well) mixed with Matrigel (1:1 ratio in 100 ml volume) were injected subcutaneously in the right and left flanks of the hormonally intact, four-week old (approximately 25 g), immune deficient (SCID), male mice (Charles River, Boston, MA). Tumour growth was monitored weekly by luciferase imaging in live animals for up to 5 weeks. Luciferin substrate (100 ml volume) was injected into mice via intraperitoneal (i.p.) route and mice were then imaged at 8-10 min post luciferin injection using Xenogen IVIS Spectrum Optical In Vivo High Resolution Bioluminescence Imaging System. Photons were counted and plotted as a function of time. Animals were sacrificed humanely and tumours were collected for morphological and molecular analysis. At 5 weeks, tumour volumes were also measured and plotted.